dconv::details Namespace Reference

Classes
struct	Power
	128 bits power. More...

Functions
constexpr bool	isDigit (char c) noexcept
constexpr bool	isSign (char c) noexcept
template<typename Probe >
bool	readExponent (Probe &probe, int64_t &exponent)
	read an optional exponent.
template<typename Probe , bool Relaxed>
DCONV_COLD bool	atodLong (Probe &probe, bool negative, uint64_t significand, int64_t &scale, bool fraction, double &value, int pending=-1)
	continue a conversion whose significand has more than 19 digits.
template<typename Probe , AtodMode Mode>
Errc	atodRelaxed (Probe &probe, double &value, uint64_t &significand, int64_t &exponent, bool &negative)
	string to double conversion in general format that stops on integers.
template<typename Probe , AtodMode Mode>
DCONV_COLD Errc	atodRewind (Probe &probe, double &value, uint64_t &significand, int64_t &exponent, bool &negative)
	parse again a number the fast path does not handle.
template<typename Probe , AtodMode Mode = AtodMode::General, bool Rewind = std::is_same_v<Probe, StringProbe>>
Errc	atod (Probe &probe, double &value, uint64_t &significand, int64_t &exponent, bool &negative)
	string to double conversion that stops on integers.
template<typename Probe , AtodMode Mode = AtodMode::General>
bool	atod (Probe &probe, double &value)
	string to double conversion.
int	countTrailingZeros (uint32_t group) noexcept
	count the trailing zeros of a four digits group.
int	countDigits (uint64_t digits) noexcept
	count the digits of a decimal significand.
int	writeDigits (char *buffer, uint64_t digits, int length) noexcept
	write the digits of a decimal significand.
char *	writeExponent (char *buffer, int k) noexcept
	write a decimal exponent.
char *	prettify (char *buffer, int length, int k) noexcept
	lay out decimal digits as a fixed or scientific number.
char *	format (char *buffer, uint64_t digits, int length, int k) noexcept
	write a decimal significand with the same layout as prettify.
void	umul128 (uint64_t a, uint64_t b, uint64_t &high, uint64_t &low) noexcept
	multiply two 64 bits unsigned integers.

Variables
constexpr uint32_t	pow10 [] = {1, 10, 100, 1000, 10000, 100000, 1000000, 10000000, 100000000, 1000000000}
	powers of ten that fit in 32 bits.
constexpr char	digitPairs []
	two digits decimal representations of 0 to 99.
constexpr uint8_t	pairZeros []
	trailing zeros of the two digits decimal representations of 0 to 99.
constexpr Power	eisellemirepow []
constexpr DiyFp	grisupow []
	cached powers of ten for the grisu algorithms.
constexpr Power	schubfachpow []

Function Documentation

atod() [1/2]

template<typename Probe , AtodMode Mode = AtodMode::General>

bool dconv::details::atod ( Probe & probe, double & value )

inline

string to double conversion.

Parameters

probe	probe to read from.
value	converted value.

Returns

true on success, false otherwise.

atod() [2/2]

template<typename Probe , AtodMode Mode = AtodMode::General, bool Rewind = std::is_same_v<Probe, StringProbe>>

Errc dconv::details::atod ( Probe & probe, double & value, uint64_t & significand, int64_t & exponent, bool & negative )

inline

string to double conversion that stops on integers.

Parameters

probe	probe to read from.
value	converted value.
significand	integer value when the error is NotDouble, first significant digits read otherwise.
exponent	decimal exponent read.
negative	sign read.

Returns

Errc::Success on success, the error otherwise.

atodLong()

template<typename Probe , bool Relaxed>

DCONV_COLD bool dconv::details::atodLong	(	Probe &	probe,
		bool	negative,
		uint64_t	significand,
		int64_t &	scale,
		bool	fraction,
		double &	value,
		int	pending = -1 )

continue a conversion whose significand has more than 19 digits.

Parameters

probe	probe to read from, positioned on the first digit beyond the significand.
negative	sign of the number.
significand	first 19 significant digits.
scale	power of ten applied to the significand, updated with the final exponent.
fraction	true if the reading stopped in the fractional part.
value	converted value.
pending	integer digit already read beyond the significand, or -1.

Returns

true on success, false otherwise.

atodRelaxed()

template<typename Probe , AtodMode Mode>

Errc dconv::details::atodRelaxed ( Probe & probe, double & value, uint64_t & significand, int64_t & exponent, bool & negative )

inline

string to double conversion in general format that stops on integers.

Parameters

probe	probe to read from.
value	converted value.
significand	integer value when the error is NotDouble, first significant digits read otherwise.
exponent	decimal exponent read.
negative	sign read.

Returns

Errc::Success on success, the error otherwise.

atodRewind()

template<typename Probe , AtodMode Mode>

DCONV_COLD Errc dconv::details::atodRewind	(	Probe &	probe,
		double &	value,
		uint64_t &	significand,
		int64_t &	exponent,
		bool &	negative )

parse again a number the fast path does not handle.

Parameters

probe	probe to read from, positioned on the start of the number.
value	converted value.
significand	integer value when the error is NotDouble, first significant digits read otherwise.
exponent	decimal exponent read.
negative	sign read.

Returns

Errc::Success on success, the error otherwise.

countDigits()

int dconv::details::countDigits ( uint64_t digits )

inlinenoexcept

count the digits of a decimal significand.

Parameters

digits

decimal significand.

Returns

number of digits.

countTrailingZeros()

int dconv::details::countTrailingZeros ( uint32_t group )

inlinenoexcept

count the trailing zeros of a four digits group.

Parameters

group

four digits group.

Returns

number of trailing zeros.

format()

char * dconv::details::format ( char * buffer, uint64_t digits, int length, int k )

inlinenoexcept

write a decimal significand with the same layout as prettify.

Parameters

buffer	buffer to write the string representation to.
digits	decimal significand.
length	number of digits.
k	decimal exponent.

Returns

end position.

isDigit()

bool dconv::details::isDigit ( char c )

inlineconstexprnoexcept

isSign()

bool dconv::details::isSign ( char c )

inlineconstexprnoexcept

prettify()

char * dconv::details::prettify ( char * buffer, int length, int k )

inlinenoexcept

lay out decimal digits as a fixed or scientific number.

Parameters

buffer	buffer holding the digits.
length	number of digits.
k	decimal exponent.

Returns

end position.

readExponent()

template<typename Probe >

bool dconv::details::readExponent ( Probe & probe, int64_t & exponent )

inline

read an optional exponent.

Parameters

probe	probe to read from.
exponent	read exponent.

Returns

true on success, false otherwise.

umul128()

void dconv::details::umul128 ( uint64_t a, uint64_t b, uint64_t & high, uint64_t & low )

inlinenoexcept

multiply two 64 bits unsigned integers.

Parameters

a	multiplicand.
b	multiplier.
high	high 64 bits of the 128-bit result.
low	low 64 bits of the 128-bit result.

writeDigits()

int dconv::details::writeDigits ( char * buffer, uint64_t digits, int length )

inlinenoexcept

write the digits of a decimal significand.

Parameters

buffer	buffer to write the digits to.
digits	decimal significand.
length	number of digits.

Returns

number of trailing zeros written.

writeExponent()

char * dconv::details::writeExponent ( char * buffer, int k )

inlinenoexcept

write a decimal exponent.

Parameters

buffer	buffer to write the exponent to.
k	exponent.

Returns

end position.

Variable Documentation

digitPairs

char dconv::details::digitPairs[]

constexpr

Initial value:

=
            "0001020304050607080910111213141516171819202122232425262728293031323334353637383940414243444546474849505152"
            "5354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899"

two digits decimal representations of 0 to 99.

eisellemirepow

Power dconv::details::eisellemirepow[]

constexpr

grisupow

DiyFp dconv::details::grisupow[]

constexpr

cached powers of ten for the grisu algorithms.

pairZeros

uint8_t dconv::details::pairZeros[]

constexpr

Initial value:

= {
            2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0,
            0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
            0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0}

trailing zeros of the two digits decimal representations of 0 to 99.

pow10

uint32_t dconv::details::pow10[] = {1, 10, 100, 1000, 10000, 100000, 1000000, 10000000, 100000000, 1000000000}

constexpr

powers of ten that fit in 32 bits.

schubfachpow

Power dconv::details::schubfachpow[]

constexpr